hpath-0.9.2: Support for well-typed paths
Copyright© 2016 Julian Ospald
LicenseBSD3
MaintainerJulian Ospald <hasufell@posteo.de>
Stabilityexperimental
Portabilityportable
Safe HaskellNone
LanguageHaskell2010

HPath.IO

Description

This module provides high-level IO related file operations like copy, delete, move and so on. It only operates on Path x which guarantees us well-typed paths. Passing in Path Abs to any of these functions generally increases safety. Passing Path Rel may trigger looking up the current directory via getcwd in some cases where it cannot be avoided.

Some functions are just path-safe wrappers around unix functions, others have stricter exception handling and some implement functionality that doesn't have a unix counterpart (like copyDirRecursive).

Some of these operations are due to their nature not atomic, which means they may do multiple syscalls which form one context. Some of them also have to examine the filetypes explicitly before the syscalls, so a reasonable decision can be made. That means the result is undefined if another process changes that context while the non-atomic operation is still happening. However, where possible, as few syscalls as possible are used and the underlying exception handling is kept.

Note: BlockDevice, CharacterDevice, NamedPipe and Socket are ignored by some of the more high-level functions (like easyCopy). For other functions (like copyFile), the behavior on these file types is unreliable/unsafe. Check the documentation of those functions for details.

Synopsis

Types

data FileType Source #

Instances

Instances details
Eq FileType Source # 
Instance details

Defined in HPath.IO

Show FileType Source # 
Instance details

Defined in HPath.IO

data RecursiveErrorMode Source #

The error mode for recursive operations.

On FailEarly the whole operation fails immediately if any of the recursive sub-operations fail, which is sort of the default for IO operations.

On CollectFailures skips errors in the recursion and keeps on recursing. However all errors are collected in the RecursiveFailure error type, which is raised finally if there was any error. Also note that RecursiveFailure does not give any guarantees on the ordering of the collected exceptions.

Constructors

FailEarly 
CollectFailures 

data CopyMode Source #

The mode for copy and file moves. Overwrite mode is usually not very well defined, but is a convenience shortcut.

Constructors

Strict

fail if any target exists

Overwrite

overwrite targets

File copying

copyDirRecursive Source #

Arguments

:: Path b1

source dir

-> Path b2

destination (parent dirs are not automatically created)

-> CopyMode 
-> RecursiveErrorMode 
-> IO () 

Copies the contents of a directory recursively to the given destination, while preserving permissions. Does not follow symbolic links. This behaves more or less like the following, without descending into the destination if it already exists:

  cp -a /source/dir /destination/somedir

For directory contents, this will ignore any file type that is not RegularFile, SymbolicLink or Directory.

For Overwrite copy mode this does not prune destination directory contents, so the destination might contain more files than the source after the operation has completed. Permissions of existing directories are fixed.

Safety/reliability concerns:

  • not atomic
  • examines filetypes explicitly
  • an explicit check throwDestinationInSource is carried out for the top directory for basic sanity, because otherwise we might end up with an infinite copy loop... however, this operation is not carried out recursively (because it's slow)

Throws:

Throws in FailEarly RecursiveErrorMode only:

Throws in CollectFailures RecursiveErrorMode only:

Throws in Strict CopyMode only:

Note: may call getcwd (only if destination is a relative path)

recreateSymlink Source #

Arguments

:: Path b1

the old symlink file

-> Path b2

destination file

-> CopyMode 
-> IO () 

Recreate a symlink.

In Overwrite copy mode only files and empty directories are deleted.

Safety/reliability concerns:

Throws:

Throws in Strict mode only:

Throws in Overwrite mode only:

Notes:

  • calls symlink
  • calls getcwd in Overwrite mode (if destination is a relative path)

copyFile Source #

Arguments

:: Path b1

source file

-> Path b2

destination file

-> CopyMode 
-> IO () 

Copies the given regular file to the given destination. Neither follows symbolic links, nor accepts them. For "copying" symbolic links, use recreateSymlink instead.

Note that this is still sort of a low-level function and doesn't examine file types. For a more high-level version, use easyCopy instead.

In Overwrite copy mode only overwrites actual files, not directories. In Strict mode the destination file must not exist.

Safety/reliability concerns:

  • Overwrite mode is not atomic
  • when used on CharacterDevice, reads the "contents" and copies them to a regular file, which might take indefinitely
  • when used on BlockDevice, may either read the "contents" and copy them to a regular file (potentially hanging indefinitely) or may create a regular empty destination file
  • when used on NamedPipe, will hang indefinitely

Throws:

Throws in Strict mode only:

Notes:

  • may call getcwd in Overwrite mode (if destination is a relative path)

easyCopy :: Path b1 -> Path b2 -> CopyMode -> RecursiveErrorMode -> IO () Source #

Copies a regular file, directory or symbolic link. In case of a symbolic link it is just recreated, even if it points to a directory. Any other file type is ignored.

Safety/reliability concerns:

Note: may call getcwd in Overwrite mode (if destination is a relative path)

File deletion

deleteFile :: Path b -> IO () Source #

Deletes the given file. Raises eISDIR if run on a directory. Does not follow symbolic links.

Throws:

deleteDir :: Path b -> IO () Source #

Deletes the given directory, which must be empty, never symlinks.

Throws:

Notes: calls rmdir

deleteDirRecursive :: Path b -> IO () Source #

Deletes the given directory recursively. Does not follow symbolic links. Tries deleteDir first before attemtping a recursive deletion.

On directory contents this behaves like easyDelete and thus will ignore any file type that is not RegularFile, SymbolicLink or Directory.

Safety/reliability concerns:

  • not atomic
  • examines filetypes explicitly

Throws:

easyDelete :: Path b -> IO () Source #

Deletes a file, directory or symlink. In case of directory, performs recursive deletion. In case of a symlink, the symlink file is deleted. Any other file type is ignored.

Safety/reliability concerns:

File opening

openFile :: Path b -> IO ProcessID Source #

Opens a file appropriately by invoking xdg-open. The file type is not checked. This forks a process.

executeFile Source #

Arguments

:: Path b

program

-> [ByteString]

arguments

-> IO ProcessID 

Executes a program with the given arguments. This forks a process.

File creation

createRegularFile :: FileMode -> Path b -> IO () Source #

Create an empty regular file at the given directory with the given filename.

Throws:

createDir :: FileMode -> Path b -> IO () Source #

Create an empty directory at the given directory with the given filename.

Throws:

createDirRecursive :: FileMode -> Path b -> IO () Source #

Create an empty directory at the given directory with the given filename. All parent directories are created with the same filemode. This basically behaves like:

  mkdir -p /some/dir

Safety/reliability concerns:

  • not atomic

Throws:

  • PermissionDenied if any part of the path components do not exist and cannot be written to
  • AlreadyExists if destination already exists and is not a directory

Note: calls getcwd if the input path is a relative path

createSymlink Source #

Arguments

:: Path b

destination file

-> ByteString

path the symlink points to

-> IO () 

Create a symlink.

Throws:

Note: calls symlink

File renaming/moving

renameFile :: Path b1 -> Path b2 -> IO () Source #

Rename a given file with the provided filename. Destination and source must be on the same device, otherwise eXDEV will be raised.

Does not follow symbolic links, but renames the symbolic link file.

Safety/reliability concerns:

  • has a separate set of exception handling, apart from the syscall

Throws:

Note: calls rename (but does not allow to rename over existing files)

moveFile Source #

Arguments

:: Path b1

file to move

-> Path b2

destination

-> CopyMode 
-> IO () 

Move a file. This also works across devices by copy-delete fallback. And also works on directories.

Does not follow symbolic links, but renames the symbolic link file.

Safety/reliability concerns:

Throws:

Throws in Strict mode only:

Notes:

  • calls rename (but does not allow to rename over existing files)
  • calls getcwd in Overwrite mode if destination is a relative path

File reading

readFile :: Path b -> IO ByteString Source #

Read the given file at once into memory as a strict ByteString. Symbolic links are followed, no sanity checks on file size or file type. File must exist.

Note: the size of the file is determined in advance, as to only have one allocation.

Safety/reliability concerns:

  • since amount of bytes to read is determined in advance, the file might be read partially only if something else is appending to it while reading
  • the whole file is read into memory!

Throws:

readFileEOF :: Path b -> IO ByteString Source #

Read the given file in chunks of size `8192` into memory until fread returns 0. Returns a lazy ByteString, because it uses Builders under the hood.

Safety/reliability concerns:

  • the whole file is read into memory!

Throws:

File writing

writeFile :: Path b -> ByteString -> IO () Source #

Write a given ByteString to a file, truncating the file beforehand. The file must exist. Follows symlinks.

Throws:

appendFile :: Path b -> ByteString -> IO () Source #

Append a given ByteString to a file. The file must exist. Follows symlinks.

Throws:

File permissions

newFilePerms :: FileMode Source #

Default permissions for a new file.

newDirPerms :: FileMode Source #

Default permissions for a new directory.

Directory reading

getDirsFiles Source #

Arguments

:: Path b

dir to read

-> IO [Path b] 

Gets all filenames of the given directory. This excludes "." and "..". This version does not follow symbolic links.

The contents are not sorted and there is no guarantee on the ordering.

Throws:

Filetype operations

getFileType :: Path b -> IO FileType Source #

Get the file type of the file located at the given path. Does not follow symbolic links.

Throws:

Others

canonicalizePath :: Path b -> IO (Path Abs) Source #

Applies realpath on the given path.

Throws:

toAbs :: Path b -> IO (Path Abs) Source #

Converts any path to an absolute path. This is done in the following way:

  • if the path is already an absolute one, just return it
  • if it's a relative path, prepend the current directory to it